Machine learning methods are usually grouped by what kind of feedback they learn from.
Supervised Learning
The model learns from labelled examples: inputs paired with the correct answer. Predicting house prices from features (regression) or deciding whether a transaction is fraudulent (classification) are supervised problems. It is the most widely used form of machine learning in business, but it depends on having good labels, which can be expensive to create.
Unsupervised Learning
The model receives unlabelled data and looks for structure on its own. Common tasks include clustering (grouping similar customers), dimensionality reduction (compressing many features into a few) and anomaly detection (spotting unusual records). The results need human interpretation, because nothing tells the algorithm what a "good" grouping is.
Reinforcement Learning
An agent takes actions in an environment and receives rewards or penalties, learning a strategy that maximises long-term reward. It powered game-playing systems such as AlphaGo and is used in robotics and control. It is also used to fine-tune language models from human feedback (RLHF).
Self-Supervised Learning
Large language models are trained with a variant: the labels come from the data itself, such as predicting the next word in a sentence. This lets models learn from huge amounts of raw text without manual labelling, and is why they can be pretrained so broadly.
Which One Do You Need?
If you have past examples with known outcomes, start with supervised learning. If you want to explore or segment data, try unsupervised methods. Reinforcement learning is powerful but harder to apply outside simulations.